Trase raises $107M seed to scale AI agents inside healthcare back offices

Interesting mostly because of where the AI is deployed. The back office — faxes, prior authorization, referral triage — is where most U.S. health systems are already spending on labor arbitrage, and where the boring wins are compounding faster than the flashier bedside pilots.

The Duke cardiology data point is worth pulling apart. 5,000+ faxes per month is a specific, quantifiable pain point that anyone who has worked in a large clinic operations team will recognize. Trase’s agent handles routing 7.1x faster than manual sorting, reclaiming approximately $285K/year of medical-assistant and nurse capacity. What Duke chose to do with that capacity is arguably more important than the capacity itself: Duke publicly stated it is redeploying the time to patient-facing work rather than headcount reduction. That framing matters because it side-steps one of the recurring political objections to health-system AI (“AI is going to fire nurses”) while still capturing the operational upside.

The $107M seed round — led by ARCH Venture Partners — is unusually large for a seed by any standard, and reflects the pattern we noted alongside Assort’s $120M Series C: capital is concentrating on infrastructure-tier “agents-in-the-back-office” plays with clear operational ROI, rather than on generic “AI copilot” tools for individual clinicians.

Watch for the Duke case study to become the reference deployment for other academic medical centers evaluating agentic back-office AI. Cardiology has an unusually long tail of inbound documents (referrals, external reports, prior auth); the pattern likely generalizes.

Related coverage: payer AI topic, health-system copilots topic.